Multi step forecasting of the wind generation by ARIMA and k-NN

Brazil is the country with the largest production of wind energy in Latin America. Moreover, it experienced a high growth in 2016, especially in its northeast region. However, wind generation has the disadvantage of intermittency. Therefore, one way to mitigate the impacts of this problem in the electric system operation is by predicting the wind speed. For this purpose, in this work were applied two different methodologies, autoregressive integrated moving average and another based in the k-nearest neighbors technique. From the predicted wind speed it was possible to estimate the energy produced by wind turbines. The results presented better accuracies with k-NN, ranging from 7,37% to 28,28% for different steps ahead. An accurate prediction of the wind speed allows a prior knowledge of the amount of energy that will be inserted in the electric power system, contributing to the system stability.